Deep Co-Space: Sample Mining Across Feature Transformation for Semi-Supervised Learning

TL;DR

Deep Co-Space is a semi-supervised learning framework that incrementally propagates labels from labeled to unlabeled images through successive feature transformations, improving classification accuracy on large-scale datasets.

Contribution

It introduces a novel incremental semi-supervised learning paradigm that selects reliable unlabeled samples based on feature transformation stability across neighborhoods.

Findings

Effective in mining unlabeled data for improved classification.

Achieves superior results on Caltech-256 and SUN-397 benchmarks.

Enhances semi-supervised learning performance with a new sample mining criterion.

Abstract

Aiming at improving performance of visual classification in a cost-effective manner, this paper proposes an incremental semi-supervised learning paradigm called Deep Co-Space (DCS). Unlike many conventional semi-supervised learning methods usually performing within a fixed feature space, our DCS gradually propagates information from labeled samples to unlabeled ones along with deep feature learning. We regard deep feature learning as a series of steps pursuing feature transformation, i.e., projecting the samples from a previous space into a new one, which tends to select the reliable unlabeled samples with respect to this setting. Specifically, for each unlabeled image instance, we measure its reliability by calculating the category variations of feature transformation from two different neighborhood variation perspectives, and merged them into an unified sample mining criterion deriving from Hellinger distance. Then, those samples keeping stable correlation to their neighboring samples (i.e., having small category variation in distribution) across the successive feature space transformation, are automatically received labels and incorporated into the model for incrementally training in terms of classification. Our extensive experiments on standard image classification benchmarks (e.g., Caltech-256 and SUN-397) demonstrate that the proposed framework is capable of effectively mining from large-scale unlabeled images, which boosts image classification performance and achieves promising results compared to other semi-supervised learning methods.